Issues with the Use of the Hampel Filter for the Recovery of Event-Related Dynamics from Electrophysiological Signals

Frequency-domain Hampel filter can eliminate narrow-band artifacts contaminating electrophysiological signals like EEG/MEG. Artifact removal is expected to unmask the underlying Event-related dynamics (ERD) and help us better understand the neural mechanisms. Though Hampel filtering has been shown to recover resting state EEG well, Hampel filter can attenuate the ERD. In this work, we investigate the effect of Hampel filter parameter tuning on the recovery/attenuation of event-related dynamics in the presence of narrow-band artifacts by simulating a gamma-band ERD masked by a synthetic Deep brain stimulation (DBS) artifact. Hampel filter was deployed on the simulated signal created by adding synthetic DBS artifact to the event-related gamma activity. We analysed the recovery of underlying gamma-ERD by the Hampel filter while varying the two filter parameters viz., window length and threshold. We determine the optimal parameter settings for recovery of event-related dynamics. We show that the ERD recovery is better with shorter window lengths and laxer thresholds but the approach results in residual artifact. In contrast, longer window lengths and stringent thresholds remove both ERD and artifact. Our results indicate that Hampel parameters be set based on the study's objectives with a longer window length if the elimination of artifacts is a priority and a shorter window length if the recovery of ERD is essential.